Cancer drug resistance: Old concept, novel solutions required

In the early days of systemic cancer treatment, clinical oncologists had few options but to use chemotherapy. Most sensitive to cytotoxic chemotherapy are dividing cells, whether of normal or tumor origin. In the past half century, we have made substantial progress in identifying the molecular drivers of cancer cells as well as in delineating the molecular tentacles by which they operate. The uncovering of both these direct culprits and indirect mediators of oncogenic transformation has guided the pharmaceutical industry in their efforts to develop sophisticated agents, which target the tumors with remarkable precision.

However, while we are increasing clinical exploitation of these advances, also with the most recent generation of precision medicines we are confronted with its hostile imperfection: drug resistance, which limits complete clinical responses. In this special Review issue of Molecular Oncology on drug resistance, several leading laboratories share their thoughts on the perspectives and challenges of precision medicines. They will do so particularly in the context of this major clinical problem: primary and acquired cancer drug resistance. The molecular mechanisms accounting for this complex issue will be discussed, along with potential solutions, for several cancer type and therapy settings.

In this foreword, I will try to briefly put some of the most significant developments in precision medicine and challenges relating to drug resistance into perspective. We have come to realize that cancer cells are addicted to certain altered genes and we have managed to translate this knowledge into vulnerabilities that we can target in patients, even when the tumor is at an advanced stage. The foundation for this perspective has been laid in the past decades, during which we have identified a few hundred cancer‐causing genes. These driver genes correspond to either oncogenes or tumor suppressor genes, which are mutated or deregulated. As a result, tumors may become dependent on these alterations for their survival and proliferation. The relevance of these drivers is illustrated also by the fact that cancers are increasingly classified in terms of the ensemble of mutant genes they harbor; targeting the proteins they encode with specific inhibitors has profound clinical impact.

The identification of genetically altered tumor drivers has spurred the development of as yet only a few dozen clinically approved precision medicines that eradicate tumor cells relatively specifically. A prime example is imatinib, a kinase inhibitor targeting the oncogene product driving chronic myeloid leukemia, BCR‐ABL, as well as mutant KIT in gastrointestinal stromal tumors. Other examples that will be discussed in this Review series are monoclonal antibodies and small molecule inhibitors targeting receptor tyrosine kinases (RTKs) and other protein kinases, as well as antibodies blocking immune checkpoints.

But we have discovered that in addition to mutant or overexpressed cancer drivers, also certain unaltered proteins are endowed with critical responsibilities in tumor cells. Among this class of what can be referred to as non‐oncogene dependencies or enablers are topoisomerases, which can be inactivated by specific inhibitors. Their effectiveness builds on a therapeutic window relying on the common incapability of tumor cells to repair their DNA adequately as well as on their lack of salvage strategies to exit the cell cycle for creating the time that is needed for this repair. Other examples include PARP, whose inactivation acts cytotoxically specifically in homologous recombination‐deficient tumors and MEK, which represents a druggable liability in tumors carrying mutations in upstream oncogene products like BRAF.

With the ability of pharmacologically targeting both drivers and enablers, nowadays clinicians have an as yet limited but steadily increasing number of precision medicines at their disposal. As we will see in this issue of Molecular Oncology, in spite of these advances, cancers that are being treated with single targeted agents are often either intrinsically resistant or acquire resistance during the course of treatment, not seldom within the first six months. Of note, the term drug resistance is often ill‐used, as it actually denotes that the therapeutic window has got lost (Borst, 2012). It goes without saying that these phenomena severely limit clinical benefit. Drug resistance occurs through a plethora of molecular mechanisms, including the acquisition of mutations blocking drug binding in the targeted proteins, induction of upstream or downstream factors through adaptation or acquisition of secondary mutations, or disruption of feedback controls.

For example, although we have seen improvements in the outcome of patients with lung cancer treated with EGFR inhibitors, a considerable number of tumors show primary resistance, while for others the duration of response does not extend beyond one year. As described by Marais and coworkers in this issue (Girotti et al., 2014), such problems are seen also in patients with melanomas harboring a BRAF mutation: while ∼15% of patients are intrinsically resistant to BRAF inhibitors, the remainder show profound initial responses, but relapse with drug‐resistant disease within the first year. Similarly, whereas BRAF mutation predisposes melanoma to elimination by the corresponding inhibitor, this is not seen in colorectal cancer, owing to feedback activation of EGFR. Reactivation of signaling pathways is commonly seen in resistance to single agent therapy. Groenendijk and Bernards point out in this issue that they had a déjà vu: these mechanisms were noted already when inhibitors of estrogen signaling, the first targeted cancer drugs, were administered to breast cancer patients (Groenendijk and Bernards, 2014). To develop more durable responses, they argue, it will be imperative to better predict and monitor such feedback mechanisms.

Given the large spectrum of (epi)genetic mechanisms underlying resistance to targeted therapy, it will also be important to develop preclinical (animal) models, to study its molecular determinants as well as to identify novel druggable factors. A major challenge is exactly that: to capture the complexity of the tumor within the context of the experimental model. Although emerging evidence demonstrates the utility of patient‐derived xenograft (PDX) models for preclinical studies, it has also become clear that biopsies when transplanted in immunodeficient mice may harbor the genetic complexity of the tumor incompletely. Therefore, Bardelli and co‐workers make a case for the use of longitudinal non‐invasive DNA profiling in patients on treatment, in parallel to using xenopatient platforms to assist in guiding therapies for patients who have failed on standard therapies (Van Emburgh et al., 2014).

The availability of next generation sequencing approaches has revolutionized cancer diagnostics, as well as basic and translational cancer research. By tumor DNA and RNA sequencing, we have uncovered a large number of (epi)genetic changes in cancer. These analyses have revealed the manifestation of profound genomic heterogeneity across cancer genomes. Thanks to the ability to perform single‐cell sequencing, we have learned that this heterogeneity is seen not only in tumors from different patients, but also in intra‐patient tumor clones and indeed within tumors. As Burrell and Swanton will discuss (Burrell and Swanton, 2014), polyclonal resistance resulting from intratumor heterogeneity poses a major clinical problem requiring advanced personalized solutions.

Sizeable international consortia have been put together to build genomic and transcriptomic databases, including ICGC and TCGA. Increasingly, publicly available resources can be mined for all kinds of (epi)genetic alterations identified in cancer cells. This notwithstanding, a significant gap remains between this wealth of data and the therapeutic options for today's clinical oncologist. As Alifrangis and McDermott will argue (Alifrangis and McDermott, 2014), among the efforts to narrow this space, one ought to perform function‐based approaches connecting the (epi)genetic makeup of the cancer cells with their spectrum of response to targeted therapies. This analysis should include non‐coding DNA stretches, accounting for the majority of the genome.

While there are already many hurdles to be taken aiming for developing more durable responses to targeted therapies in general, in the case of particularly brain metastases yet another layer of complexity limits effective treatment. Several tumor types, most prominently lung cancer, have a propensity to spread to the brain, severely impacting on prognosis and quality of life. As Seoane and De Mattos‐Arruda will explain (Seoane and De Mattos‐Arruda, 2014), in addition to therapy resistance, brain metastases are associated with complex pathologies, including the blood–brain barrier and transporters therein preventing drugs from reaching the tumor and micrometastases. These and other factors, such as poor accessibility to both systemic therapies and immune cells, cause therapies that are relatively successful for extracranial tumors to fail in this setting.

Amidst the relative success of agents targeting cancer cell‐intrinsic vulnerabilities, we have seen in recent years another encouraging development, namely, immunotherapy. Many decades of basic research into the role and regulation of the (adaptive) immune system have led to the identification of several factors that together constitute immune ‘checkpoints’. Recent clinical studies have revealed that monoclonal antibodies targeting such factors, most prominently CTLA4 and PD‐1, are exceedingly effective in suppressing tumor growth. This has been observed for metastatic melanoma, which has become a paradigm tumor type for the development of both targeted and immunotherapies, but is currently also being explored for other tumor types. As Kelderman, Schumacher and Haanen, and also Marais will discuss here (Kelderman et al., 2014; Girotti et al., 2014), it is too early to put a number on it, but at least in the context of melanoma, durable complete responses may soon no longer be exceptions. But also for immunotherapy, primary and acquired resistance continue to limit the effectiveness of the treatment. Little is known about the underlying mechanisms to date, but many efforts are ongoing to crack also this nut.

These and other emerging aspects of drug resistance will be discussed in this special Review issue, covering primary tumors and metastases, targeted and immunotherapies, and intrinsic and acquired resistance mechanisms. We have come a long way, from uncovering the drivers and enablers in cancer cells to identifying among them the factors that are amenable to therapeutic intervention. And our efforts are beginning to pay off. Despite this major progress, unfortunately, we have learned that cancer cells utilize hard‐ and soft‐wired programs of plasticity. Thus, as for chemotherapy, resistance to targeted agents represents a formidable roadblock to long‐lasting clinical responses.

With advanced genetic analyses, we have begun to discover also the molecular hallmarks of drug resistance mechanisms at a rapid pace. To tackle the problem, basic and translational cancer researchers, bioinformaticians and clinical oncologists will need to join hands with proteomic, imaging, metabolomic and animal model experts. They will collect complementary pieces of information on the genetic composition of the tumor, its longitudinal response to treatment, its heterogeneity and other clinically relevant biomarkers and parameters. This will lay the foundation for developing clinical strategies that overcome, or better: prevent, drug resistance. It is conceivable that combinatorial therapies comprising precision agents exploiting both cancer cell‐intrinsic and ‐extrinsic vulnerabilities will contribute to solving this dire clinical problem.

Peeper Daniel S., (2014), Cancer drug resistance: Old concept, novel solutions required, Molecular Oncology, 8, doi: 10.1016/j.molonc.2014.07.026.